Fix/clip post-lapse stability

[L-M-Sherlock]

related PR:

• Fix/consider short term params when clipping post-lapse stability fsrs-rs#257

[ishiko732]

When the long-term scheduler is enabled, should $w_{17} \times w_{18}$ be equal to 0?

I currently only know that $w_{17}$ and $w_{18}$ are related to next_short_term_stability, which is not called by the long-term scheduler.

[L-M-Sherlock]

When the long-term scheduler is enabled, should
𝑤
17
×
𝑤
18
be equal to 0?

Yes. So exp(w[17] * w[18]) = 1.

[ishiko732]

So exp(w[17] * w[18]) = 1.

I understand, but after switching to the long-term scheduler, $$w_{17}$$ and $$w_{18}$$ are not necessarily 0. Should their values be determined based on p.EnableShortTerm?

var newSMin float64
if p.EnableShortTerm {
    newSMin = s / math.Exp(p.W[17]*p.W[18])
} else {
    newSMin = s
}

// or
var newSMin float64 = s
if p.EnableShortTerm {
    newSMin = s / math.Exp(p.W[17]*p.W[18])
}

[L-M-Sherlock]

Right. I have another idea: when EnableShortTerm is false, reset W[17] and W[18] to zero. What about that?

[ishiko732]

I have another idea: when EnableShortTerm is false, reset W[17] and W[18] to zero. What about that?

I think that continuously switching schedulers might confuse developers, as they may not understand why W[17] and W[18] are suddenly set to zero. :(

[L-M-Sherlock]

OK, I will update the PR based on your suggestion tomorrow.

[L-M-Sherlock]

@ishiko732 cc

[ishiko732]

LGTM